Re: [Hyperic-hq-discuss] cpu-counters on linux/ia64

[Doug M. <do...@hy...>]

Many thanks Jan, I have committed the change to trunk and branches/=20
SIGAR_1_3

On Mar 29, 2007, at 12:28 PM, Jan Kneschke wrote:

> Hi Doug,
>
> I get
>
> sigar-test-all.c.98 (unknown):
>   cpu.user =3D 0,
>   cpu.sys =3D 0,
>   cpu.nice =3D 0,
>   cpu.idle =3D 0,
>   cpu.wait =3D 0,
>   cpu.total =3D 0
>
> on linux/ia64 (2.4.21-37.EL), 2cpu:
>
> [0]
>   info.vendor =3D Intel,
>   info.model =3D Itanium 2,
>   info.mhz =3D 1299,
>   info.cache_size =3D 0
>
> The problem with the counters is:
>
> #define SIGAR_TICK2MSEC(s) ((sigar_uint64_t)(s) * \
>   ((sigar_uint64_t)SIGAR_MSEC / sigar->ticks))
>
> SIGAR_MSEC   =3D 1000
> sigar->ticks =3D 1024
>
> SIGAR_MSEC / sigar->ticks =3D 0
>
> This turns all the numbers into 0. This is a integer-division.
>
> Typecasting to double turns it into a floating-point division can =20
> fixes
> the problem:
>
> /* cpu ticks to milliseconds */
> #define SIGAR_TICK2MSEC(s) (sigar_uint64_t)((sigar_uint64_t)(s) *
> ((sigar_uint64_t)SIGAR_MSEC / (double)sigar->ticks))
> #define SIGAR_TICK2NSEC(s) (sigar_uint64_t)((sigar_uint64_t)(s) *
> ((sigar_uint64_t)SIGAR_NSEC / (double)sigar->ticks))
>
> Jan
> --
> Jan Kneschke, MySQL GmbH, Radlkoferstr. 2, D-81373 M=FCnchen
> Gesch=E4ftsf=FChrer: Hans von Bell, Kaj Arn=F6 - HRB M=FCnchen 162140
>
>